Hostname: page-component-7b7567c8c-fvjsw Total loading time: 0 Render date: 2026-11-14T00:10:54.905Z Has data issue: false hasContentIssue false
  • English
  • Français

From bottlenecks to better practices: systemwide electronic clinical decision support for blood culture stewardship

Published online by Cambridge University Press:  27 July 2026

Jessica SeidelmanOpen the ORCID record for Jessica Seidelman[Opens in a new window],
Michael YarringtonOpen the ORCID record for Michael Yarrington[Opens in a new window],
Stefani Dubea and
Jonathon Bae
Jessica Seidelman*
Affiliation:
Infectious Diseases, Duke University Hospital, Durham, USA
Michael Yarrington
Affiliation:
Duke University Health System, USA
Stefani Dubea
Affiliation:
Duke University Hospital, USA
Jonathon Bae
Affiliation:
Duke University Health System, USA
*
Corresponding author: Jessica Seidelman; Email: jessica.seidelman@duke.edu
Rights & Permissions [Opens in a new window]

Abstract

A systemwide electronic clinical decision support (CDS) tool for blood culture ordering reduced utilization by 12% across three hospitals (IRR 0.88, P < .01), but the effect waned over time. Low-yield ordering persisted, and contamination rates increased. Sustained diagnostic stewardship likely requires multimodal strategies beyond CDS alone.

Information

Type
Concise Communication
Information
Infection Control & Hospital Epidemiology , Volume 47 , Issue 9 , September 2026, pp. 959 - 961
Check for updates
Copyright
© The Author(s), 2026. Published by Cambridge University Press on behalf of The Society for Healthcare Epidemiology of America

Introduction

Blood culture overuse contributes to unnecessary antibiotic exposure, false-positive results due to contamination, and increased healthcare costs.Reference Fabre, Klein and Salinas1–Reference Woods-Hill, Colantuoni and Koontz4 Diagnostic stewardship interventions, including clinical decision support (CDS), have been proposed to improve appropriateness of blood culture utilization.Reference Fabre, Davis and Diekema5 In August 2024, a national shortage of blood culture bottlesReference Itoh, Akazawa-Kai, Okumura, Kuriki, Wachino and Kawabata6 prompted implementation of a systemwide CDS tool embedded in the electronic medical record (EMR) to encourage evidence-based blood culture ordering.

We evaluated the impact of this CDS intervention on blood culture event (BCE) rates and downstream clinical outcomes. We hypothesized that CDS would reduce BCE rates without adversely affecting patient outcomes.

Methods

This quasi-experimental study was conducted across a three-hospital academic health system. The intervention, implemented in August 2024, consisted of EMR-integrated CDS triggered whenever a provider ordered a blood culture set. We only included patients in the inpatient setting and excluded patients seen in the outpatient setting and the emergency department.

Providers were encouraged to use a blood culturing ordering panel in our EMR. The ordering panel first asks providers to specify how many blood cultures they want to order and the source, highlighting that the strong preference is for two peripheral blood cultures. The next screen then presents a two-sided panel with the left-side outlining scenarios that are high risk, medium risk, and low risk of bacteremia, which was adapted from the DISTRIBUTE study algorithm (Supplemental Figure 1).Reference Fabre, Klein and Salinas1 On the right side of the panel providers are then first asked if the patient meets criteria for sepsis. If “yes” is selected, then no additional information is required prior to signing the blood culture orders. However, if “no” is selected to the sepsis-related question, then the provider is prompted to select the clinical indication for blood cultures. For low-yield indications, CDS asked the provider to please document the clinical reasoning for ordering blood cultures with an indication that has low risk for bacteremia. The provider could either select “Immunocompromised” or “Other.” If “Other” was selected, a comment box was displayed for a free text response. Providers retained the ability to proceed with ordering.

A BCE was defined as collection of a blood culture set or blood culture sets ordered by a clinician for a specific clinical indication. For example, two sets of blood cultures drawn for a fever on hospital day 3 were counted as one BCE.

The preintervention period was August 2022–July 2024, and the postintervention period was August 2024–July 2025. The primary outcome was BCE rate per 1,000 inpatient days. Secondary outcomes included in-hospital mortality, 30-day readmission, blood culture contamination, antibiotic utilization (days of therapy per 1,000 inpatient days), central-line-associated bloodstream infection (CLABSI) rates, and hospital-onset Clostridioides difficile infection (HO-CDI) rates.

Interrupted time series (ITS) analysis was used to evaluate changes in BCE rates over time. We used segmented Poisson regression models with a log link and offset terms for patient-days (or central venous catheter-days for CLABSI) to conduct ITS analyses. Models included terms for time, intervention implementation, and time after intervention to estimate preintervention trends, immediate level changes, and changes in postintervention slopes.

Results

BCE rates decreased from 52.8 to 46.5 per 1,000 inpatient days following CDS implementation (incidence rate ratio [IRR] 0.88, 95% CI 0.87–0.89; P < .01). ITS analysis demonstrated an immediate 17.5% (95% CI −20.7%, 014.2%, P value < .01) reduction in BCE rates at the time of intervention and a subsequent monthly increase of 1.0% (95% CI 0.6%, 1.4%, P value < .01) in BCE rates during the postintervention period. (Figure 1) During the intervention period, 37,418 BCEs were ordered, of which 61.5% were for sepsis; 22.0% of non-sepsis orders were for low-risk clinical scenarios. Among low-risk indications, the most common were isolated fever/leukocytosis (73.8%), pneumonia (8.7%), and skin and soft tissue infection (8.6%).

Figure 1.

Interrupted time series of before and after clinical decision support implementation looking at blood culture event (BCE) rates. legend: β1 indicates a .2% increase per month (95% 0.02%, 0.3%, = −0.22) in BCE rates in the preintervention stage. Following the intervention, β2 shows that the BCE rate dropped by 17.5% (95% CI −20.7%, 014.2%, P value < .01). after that β3 demonstrates that the BCE rate increased by 1.0% per month (95% CI 0.6%, 1.4%, P value< .01). Incidence rate ratio comparing pre versus postintervention BCE rates is 0.94 (95% 0.92, 0.96), P value < .01).

A line graph showing blood culture event rates per 1000 patient days over time, with different phases indicated by beta labels.

Secondary outcome ITS analyses are depicted in Figure 2. Secondary ITS analyses demonstrated no significant changes in 30-day readmission, antibiotic days of therapy, or HO-CDI following implementation. In-hospital mortality exhibited a significant declining preintervention trend that accelerated after intervention implementation (postintervention slope RR 0.98, 95% CI 0.96–0.99). Blood culture contamination demonstrated an immediate increase at intervention onset (RR 2.03, 95% CI 1.33–3.10), followed by a significant decline over time (postintervention slope RR .94, 95% CI 0.89–0.99). CLABSI rates demonstrated a significant preintervention decline but no significant postintervention change.

Figure 2.

Interrupted time series analysis for secondary outcomes of hospital mortality, 30-day all-cause readmission, blood culture contamination, antibiotic days of therapy, central-line-associated bloodstream infection, and hospital-associated Clostridium difficile infection rates during the preintervention period 8/2022–7/202 and the intervention period 8/2024–7/2025. legend: a, in-hospital mortality; b, 30-day all-cause readmission; c, blood culture contamination; d, antibiotic days of therapy; e, central-line-associated bloodstream infection; f, hospital-associated Clostridium difficile infection.

Six line graphs depict observed and predicted rates over time for various hospital outcomes.

Discussion

Implementation of EMR-based CDS was associated with an immediate reduction in blood culture utilization, supporting its effectiveness as a diagnostic stewardship tool during a resource-constrained period. However, this effect was not sustained, with BCE rates increasing over time.

Importantly, a substantial proportion of blood cultures continued to be ordered for low-yield indications, suggesting that CDS alone may be insufficient to drive durable behavior change. This aligns with prior literature indicating that passive or interruptive CDS require reinforcement through complementary strategies such as audit-and-feedback, peer comparison, or accountability structures.Reference Nanji, Seger and Slight7,Reference Nanji, Slight and Seger8

A potential weakness of this study is that providers may have known about the national blood culture shortage regardless of the CDS tool. It is difficult to determine whether BCE rates changed due to changes in provider behavior versus implementation of the CDS tool. Additionally, it is difficult to assign causality of the CDS to the secondary outcomes. Specifically, the observed increase in contamination rates and antibiotic utilization and the decrease in mortality are unlikely related to the intervention versus an unidentified confounder. These findings highlight the complexity of diagnostic stewardship and the need for multimodal interventions.

Conclusions

Systemwide CDS implementation led to a transient reduction in blood culture utilization but did not produce sustained improvements. Continued ordering in low-risk scenarios underscores the need for additional stewardship strategies beyond CDS alone. Future studies should evaluate whether integrating CDS with complementary behavioral strategies—such as real-time audit-and-feedback, peer benchmarking, and indication-specific hard stops for the lowest-yield scenarios—can sustain reductions in blood culture utilization.

Supplementary material

The supplementary material for this article can be found at https://doi.org/10.1017/ice.2026.10515.

Acknowledgments

None.

Financial support

None reported.

Competing interests

All authors report no conflicts of interest relevant to this article.

References

Fabre, V, Klein, E, Salinas, AB, et al. A diagnostic stewardship intervention to improve blood culture use among adult nonneutropenic inpatients: The DISTRIBUTE study. J Clin Microbiol 2020;58. doi: 10.1128/jcm.01053-20.CrossRefGoogle ScholarPubMed
Geisler, BP, Jilg, N, Patton, RG, Pietzsch, JB. Model to evaluate the impact of hospital-based interventions targeting false-positive blood cultures on economic and clinical outcomes. J Hosp Infect 2019;102:438–444. doi: 10.1016/j.jhin.2019.03.012.CrossRefGoogle ScholarPubMed
Schinkel, M, Boerman, A, Carroll, K, et al. Impact of blood culture contamination on antibiotic use, resource utilization, and clinical outcomes: A retrospective cohort study in Dutch and US Hospitals. Open Forum Infect Dis 2024;11:ofad644. doi: 10.1093/ofid/ofad644.CrossRefGoogle ScholarPubMed
Woods-Hill, CZ, Colantuoni, EA, Koontz, DW, et al. Association of diagnostic stewardship for blood cultures in critically ill children with culture rates, antibiotic use, and patient outcomes: Results of the Bright STAR Collaborative. JAMA Pediatr 2022;176:690–698. doi: 10.1001/jamapediatrics.2022.1024.CrossRefGoogle ScholarPubMed
Fabre, V, Davis, A, Diekema, DJ, et al. Principles of diagnostic stewardship: A practical guide from the Society for Healthcare Epidemiology of America Diagnostic Stewardship Task Force. Infect Control Hosp Epidemiol 2023;44:178–185. doi: 10.1017/ice.2023.5.CrossRefGoogle Scholar
Itoh, N, Akazawa-Kai, N, Okumura, N, Kuriki, S, Wachino, C, Kawabata, T. Impact of BD BACTEC blood culture bottle shortage on performance metrics: An interrupted time-series analysis at a Japanese university-affiliated hospital. J Infect Chemother 2025;31:102664. doi: 10.1016/j.jiac.2025.102664.CrossRefGoogle Scholar
Nanji, KC, Seger, DL, Slight, SP, et al. Medication-related clinical decision support alert overrides in inpatients. J Am Med Inform Assoc 2018;25:476–481. doi: 10.1093/jamia/ocx115.CrossRefGoogle ScholarPubMed
Nanji, KC, Slight, SP, Seger, DL, et al. Overrides of medication-related clinical decision support alerts in outpatients. J Am Med Inform Assoc 2014;21:487–91. doi: 10.1136/amiajnl-2013-001813.CrossRefGoogle ScholarPubMed
Figure 0

Figure 1.Interrupted time series of before and after clinical decision support implementation looking at blood culture event (BCE) rates. legend: β1 indicates a .2% increase per month (95% 0.02%, 0.3%, = −0.22) in BCE rates in the preintervention stage. Following the intervention, β2 shows that the BCE rate dropped by 17.5% (95% CI −20.7%, 014.2%, P value < .01). after that β3 demonstrates that the BCE rate increased by 1.0% per month (95% CI 0.6%, 1.4%, P value< .01). Incidence rate ratio comparing pre versus postintervention BCE rates is 0.94 (95% 0.92, 0.96), P value < .01).

Figure 1

Figure 2.Interrupted time series analysis for secondary outcomes of hospital mortality, 30-day all-cause readmission, blood culture contamination, antibiotic days of therapy, central-line-associated bloodstream infection, and hospital-associated Clostridium difficile infection rates during the preintervention period 8/2022–7/202 and the intervention period 8/2024–7/2025. legend: a, in-hospital mortality; b, 30-day all-cause readmission; c, blood culture contamination; d, antibiotic days of therapy; e, central-line-associated bloodstream infection; f, hospital-associated Clostridium difficile infection.

Supplementary material: File

Seidelman et al. supplementary material

Seidelman et al. supplementary material
Download Seidelman et al. supplementary material(File)
File 1.5 MB